[Mesa-dev] postmortem: a306 job fails on db410c-1 today

[Eric Anholt]

Looks like in configuring the runners to do !5661's nginx for getting
artifacts (.qpa xml files, particularly), I forgot that I had
previously used that port for passthrough caching of fd.o access to
reduce egress (which has been unused since switching off of LAVA).
And, after identifying the problem this afternoon, I turned off the
(also unused) caching nginx on the servo runner instead of the
fastboot runner, so the problem continued into the evening.  We hadn't
identified it in merging !5661 by apparently not ending up with any of
our 3 jobs allocated to #1/7 active runners in any of the pipelines
before merging.

Resolution: When enabling new runner functionality, try multiple
pipelines running in parallel to try to make sure that all the runners
get tested.